The latest Mastercard Signals report, “Generative AI: The transformation of banking”, addresses the impact of generative artificial intelligence on the financial sector. The authors recognize the potential of this technology in improving the functioning of banks; from streamlining data processing processes, facilitating staff work, to providing customers with more personalized experiences and services. At the same time, they also perceive obstacles – including concerns related to data confidentiality or misinformation.
About a year ago, the concept of generative artificial intelligence was known primarily to AI engineers and data analysts. Today, this technology has entered the lives and awareness of consumers and is at the forefront of the economic revolution. The authors of the Mastercard report cite research indicating that 55% of surveyed CEOs of large global companies confirmed that they are “evaluating or experimenting” with generative artificial intelligence, and 37% are already using it.
As of now, both banks and other financial institutions present a rather conservative approach to artificial intelligence. The first deployments based on this technology are internal – banks use them mainly to support their own solutions in the field of management systems and conducting analysis. As trust in generative AI grows, however, it has the potential to become an integral part of financial services. It can potentially changethe dynamics of competition in banking, strengthening the position of new entities entering the market or changing the balance of power among existing institutions. That’s why Mastercard analyzed the potential and challenges associated with using artificial intelligence in the banking industry.
AI Potential in Banking
The areas and examples of using artificial intelligence that are likely to appear in banking in the next five to seven years could be as follows:
Knowledge and analysis – Bankers equipped with artificial intelligence may find that searching for information, which once took hours, can now be done in minutes.
Information technology – Artificial intelligence can assist in developing project specifications: writing codes, or creating synthetic data, with which can test new solutions for potential fraud and risk assessment systems. On a daily basis, engineers can use artificial intelligence to obtain tips helpful in performing current tasks.
Cybersecurity and fraud prevention – Large language models (LLM) can be adapted to work in the field of security. They could react to threats and transform complex data into clear instructions on which specialists could take appropriate actions. The generative AI’s ability to recognize patterns can potentially improve the monitoring capabilities of older forms of this technology.
Talent management – Thanks to the ability to process unstructured data, AI-based solutions can find and present to HR managers candidates who, despite a lack of traditional banking education, can offer valuable competencies in working in a given institution or in a specific position.
Customer implementation – Artificial intelligence can streamline documentation management related to a bank’s clients. By quickly synthesizing data, it can signal potential risks and automate formalities.
Conversational banking – Artificial intelligence can assist bots, which will become capable of responding to customer inquiries in a contextually appropriate way. Today, many bank customers try to bypass the chat system to reach a consultant – this may change in the future.
Wealth advisory – Solutions based on generative artificial intelligence can provide financial advice that is not burdened with emotions or wishful thinking.
Granting loans – Artificial intelligence can shorten the time it takes to process loan applications and reduce associated costs, offering applicants step-by-step conversational directions.
Loyalty programs – Generative AI provides those managing loyalty programs with a potential tool for real-time communication with program participants about their expectations, enabling better matching to consumer needs.
Marketing and communication – In addition to using generative artificial intelligence to create emails and posts on social media, marketers can gain a new understanding of consumer reactions, combining its content-generating capabilities with sentiment analysis and trends in social media.
Minimizing Challenges
Generative artificial intelligence also brings with it unique challenges that banks must face.
Data confidentiality – To ensure privacy, banks could selectively use artificial intelligence models with closed and open source codes, using self-protection methods, starting from firewalls, through engineering protocols to data tokenization. If possible, banks could also build their own LLM from scratch.
Inaccuracy – Generative AI may be susceptible to “deceptions” and other inaccuracies. The financial sector, susceptible to negative information, must therefore take a number of preventive measures, including: designing, creating, developing and refining queries directed to AI language models so that they provide precise answers, targeted model tuning and, most importantly, human supervision.
Data integrity – Banks must ensure that the information conveyed is accurate, credible and free from errors or bias. The output data from an algorithmic system will need to be maximally identifiable, explainable and trustworthy. Traditional data management practices, LLM tuning, consultations with people and continuous audits can help banks in this area.
Data availability – Technological barriers in banks can cause data to be placed in inaccessible systems and storage environments. However, banks can expect extended API networks and a developing plugin infrastructure, which will facilitate data flow.
“Generative Artificial Intelligence will only become truly useful when we are able to trust it, and the only way to do this is to apply basic principles of accountability to it and the data used to create it. At Mastercard, artificial intelligence models support many solutions, securing over 125 billion transactions in our network every year. By employing hundreds of data analysts and AI engineers, we are committed to developing practical artificial intelligence solutions that care about privacy and ethics from the very beginning. In all our activities, we guarantee that artificial intelligence is used responsibly and ethically,” says Andrew Reiskind, Chief Data Officer at Mastercard.
The complete report is available at:
https://www.mastercard.com/news/insights/mastercard-signals/
